13 October 2023

Changes DIPlib 3.4.1

Changes to DIPlib

dip::Image::At( dip::Image mask ) will throw an exception if the mask has no pixels set. This ensures that a dip::Image::View always selects at least one pixel, it no longer is possible to create an empty View object.
dip::CopyFrom( dip::Image const& src, dip::Image& dest, dip::Image const& srcMask ) returns a raw image instead of trying to allocate space for 0 pixels if srcMask has no pixels set. Forging an image with 0 pixels is not possible.
dip::ImageRead() produces a better error message if the file doesn’t exist.
Improved error messages when reading or writing JPEG files.
dip::DrawPolygon2D() no longer requires that all vertices be within the image for the non-filled case (the filled case never required it).

dip::Image out = view threw an exception with the message “Sizes must be non-zero and no larger than 9223372036854775807” if the view was empty. This could happen if view was created using a mask with no pixels set. This issue has been corrected by making it impossible to create an empty View object.
Note that out = img.At( img < 0 ) still throws an exception if img has no negative values, because the indexing operation is now illegal. The exception thrown now, “The mask image selects no pixels”, is more useful.
The dip::FileInformation data structure crated for a NPY file reported the file type was "NYP".
Third and fourth order cubic spline interpolation (methods "3-cubic" and "4-cubic") now always use double-precision floats internally for computation. The improved precision is necessary to evaluate the third-order polynomials with sufficient precision, the errors could be seen in flat areas of the rescaled image.
dip::Sharpen() and dip::UnsharpMask() didn’t work correctly when the input and output images were the same.
dip::BesselJN( 0, 0 ) returned 0.0 instead of 1.0.
dip::BinaryPropagation unwittingly changed behavior in 3.4.0: When iterations was 0, the seed pixels outside the mask made it into the output image, but shouldn’t have. See issue #135.

None, but see changes to DIPlib.

None, but see bugfixes to DIPlib.

Significantly faster load times (PR #127, PR #128). The undocumented variable dip.hasDIPjavaio no longer exists. The DIPjavaio module is loaded when first used.
Updated the version of Bio-Formats that is downloaded with python -m diplib download_bioformats to 7.0.0.

(See also changes to DIPlib.)

None, but see bugfixes to DIPlib.

None.

Images of more than 2 GB can now be read using Bio-Formats. But each single x-y plane still must be smaller than 2 GB. See issue #133.
When reading through Bio-Formats, pixel sizes reported in micron would be ignored with a warning. See issue #133.